Interference of Blood Cell Lysis on Routine Testing

Giuseppe Lippi, MD; Martina Montagnana, MD; Gian Luca Salvagno, MD; Gian Cesare Guidi, MD

● Context.—Preanalytical factors influencing the reliability observed in activated partial thromboplastin time and fi- of laboratory testing are commonplace. It is traditionally brinogen in samples containing a final lysate concentration accepted that hemolytic samples are unsuitable for coag- of 0.9%. The current analytical quality specifications for ulation assays because of the release of , intra- desirable bias are ؎2.0% for , ؎2.3% cellular components, and thromboplastic substances from for activated partial thromboplastin time, and ؎4.8% for damaged blood cells. fibrinogen. Percent variations from the baseline values ex- Objective.—To evaluate the influence of blood cell lysis ceeding the current analytical quality specifications for de- on routine coagulation testing. sirable bias were achieved for lysate concentrations of Design.—Twelve aliquots prepared by serial dilutions of 0.9% (prothrombin time and activated partial thrombo- homologous lysated samples collected from 10 different plastin time) and 1.8% (fibrinogen), corresponding to av- subjects, and displaying a final percentage of lysis ranging erage free plasma hemoglobin concentrations of 1.7 and from 0% to 9.1%, were tested for prothrombin time, ac- 3.4 g/L, respectively. tivated partial thromboplastin time, fibrinogen, and dimer- Conclusion.—Our results confirm that, although slightly ized plasmin fragment D. Lysis was achieved by subjecting hemolyzed specimens might still be analyzable, a moder- citrated whole blood to a freeze-thaw cycle. ate blood cell lysis, as low as 0.9%, influences the reli- Outcome Measures.—Interference from blood cell lysis ability of routine coagulation testing. Because the interfer- on routine coagulation testing. ence in coagulation assays has a wide interindividual bias, Results.—Statistically significant increases in prothrom- we do not recommend lysis correction and we suggest that bin time and dimerized plasmin fragment D were observed the most appropriate corrective measure should be free in samples containing final lysate concentrations of 0.5% hemoglobin quantification and sample recollection. and 2.7% respectively, whereas significant decreases were (Arch Pathol Lab Med. 2006;130:181–184)

oagulation testing is a central aspect of the diagnostic lems arising from a cumbersome blood draw, such as un- C approach to patients with disturbances, satisfactory attempts to draw blood, difficulty locating and it is pivotal for monitoring antithrombotic therapies easy venous accesses, and missing the vein, are anecdot- with either heparins or oral anticoagulants.1,2 Among ma- ally known to produce major interference in routine co- jor determinants of coagulation testing, the standardiza- agulation assays, and the use of hemolyzed specimens has tion of the preanalytical phase exerts a major influence on been discouraged to avoid unreliable results.4,5 However, result reliability. A standardized procedure for specimen to our knowledge, little is known about the true influence collection was demonstrated to be essential to achieve ac- of unsuitable samples caused by blood cell lysis on routine curate and precise measurements, which might finally coagulation testing. Therefore, we evaluated the interfer- provide appropriate and suitable clinical information.3 ence of in vitro blood cell lysis on prothrombin time (PT), However, there are additional circumstances besides spec- activated partial thromboplastin time (aPTT), fibrinogen, imen collection that might influence the results of coagu- and dimerized plasmin fragment D (D-dimer) testing. lation testing; these additional circumstances might also generate misleading results and induce an inappropriate MATERIALS AND METHODS diagnostic or therapeutic approach to the patients. Prob- Experiment Design and Blood Sampling On the morning of the first day of the evaluation, 4.5 mL of Accepted for publication September 12, 2005. blood were separately collected in 2 siliconized vacuum tubes From the Istituto di Chimica e Microscopia Clinica, Dipartimento di containing 0.5 mL of 0.105M buffered trisodium citrate (Becton Scienze Morfologico-Biomediche, Universita` degli Studi di Verona, Ve- ϫ rona, Italy. Dickinson, Oxford, United Kingdom), using a 20-gauge, 0.80 The authors have no relevant financial interest in the products or 19-mm Venoject multisample straight needle (Terumo Europe companies described in this article. NV, Leuven, Belgium) from 10 healthy volunteers. Volunteers Corresponding author: Giuseppe Lippi, MD, Istituto di Chimica e gave an explicit informed consent for the investigation. Volunteers Microscopia Clinica, Dipartimento di Scienze Morfologico-Biomedi- were selected on the basis of a homogeneous range of values for che, Universita` degli Studi di Verona, Ospedale Policlinico G.B. Rossi, white blood cell counts (3.25–4.98 ϫ 103 cells/␮L), count Piazzale Scuro, 10-37134—Verona, Italy (e-mail: [email protected]; (153–281 ϫ 103 cells/␮L), and hemoglobin concentration (137– [email protected]). 146 g/L). The first specimen (sample 1) was gently mixed by Reprints not available from the authors. inverting the tube 4 to 6 times and immediately stored at Ϫ70ЊC, Arch Pathol Lab Med—Vol 130, February 2006 Hemolysis and Coagulation Testing—Lippi et al 181 whereas the second specimen (sample 2) was gently mixed by aPTT (Figure, B) and fibrinogen (Figure, C) values were inverting the tube 4 to 6 times and then centrifuged at 3000g for substantially decreased when compared with the baseline Њ Ϫ Њ 10 minutes at 10 C. Plasma was separated and stored at 70 C. specimens (no lysate). Statistically significant differences On the morning of the second day of the evaluation, blood was by the Student paired t test were observed in samples con- collected into 6 additional siliconized vacuum tubes containing 0.105M buffered trisodium citrate (Becton Dickinson) using a 20- taining a final lysate concentration of 0.5% for PT, 0.9% gauge, 0.80 ϫ 19-mm Venoject multisample straight needle (Ter- for aPTT and fibrinogen, and 2.7% for D-dimer. Percent umo) from each of the 10 volunteers. All samples were gently variations from the baseline value exceeding the current mixed by inverting the tube 4 to 6 times, pooled, and divided analytical quality specifications for desirable bias were into 12 aliquots of 2 mL each. Samples 1 and 2 were thawed. achieved for lysate concentrations of 0.9% (PT and aPTT) Twelve serial dilutions, obtained by mixing samples 1 and 2, and 1.8% (fibrinogen), corresponding to average free plas- were prepared by adding 200 ␮L of each dilution to the 2-mL ma hemoglobin concentrations of 1.7 and 3.4 g/L, respec- aliquots of blood collected on the second day. Final whole blood tively. There was a roughly linear relationship between the lysate concentrations in the mixtures of samples ranged from 0% degree of variation and the percentage of lysate in the to 9.1%, roughly corresponding to average free plasma hemoglo- plasma. However, an unpredictable, sample-specific re- bin concentrations ranging from 0 to 17 g/L, thus, almost rep- resentative of the degree of hemolysis that we observe in speci- sponse was observed for each of the parameters assayed, mens sent to our laboratory. The blood samples were centrifuged as shown by the amplitude of the mean coefficients of at 3000g for 10 minutes at 10ЊC, and plasma was separated and variation, which ranged from 24% to 28% for PT, from immediately analyzed. 19% to 20% for aPTT, from 23% to 27% for fibrinogen, and from 54% to 62% for D-dimer. Because the clotting assays Laboratory Testing were performed at the alternative 570-nm wavelength, Hemolysis was assayed by measuring the concentration of free these variations were almost referable to the direct effect plasma hemoglobin by the reference cyanmethemoglobin method of hemoglobin, intracellular components, and thrombo- on a UV-1700 Spectrophotometer (Shimadzu Italia S.l.r., Milan, plastic substances released by damaged cells, rather than 6 Italy). White blood cell and platelet counts were performed on to an optical interference. an ADVIA 120 (Bayer Diagnostics, Newbury, Berkshire, United Kingdom). Routine coagulation measurements were performed COMMENT on a Behring Coagulation System (Dade-Behring, Marburg, Ger- many), using proprietary reagents: Thromborel S (lyophilized Hemolysis, causing leakage of hemoglobin and other in- human placental thromboplastin), Pathromptin SL (vegetable ternal components from the erythrocyte membrane into phospholipid with micronized silica), and Multifibren U, for PT, the surrounding fluid, is usually defined for extracellular aPTT, and fibrinogen testing, respectively. Because hemolytic hemoglobin concentrations greater than 0.3 g/L (18.8 samples have a greater than usual inherent spectrophotometric mmol/L). Hemolysis confers a detectable pink-to-red hue absorbance in conventional clotting assays, PT, aPTT, and fibrin- to serum or plasma and becomes clearly visible in speci- ogen measurements were performed at a different wavelength mens containing as low as 0.5% hemolysate.8 Hemolyzed (570 nm), uniquely available on the BCS analyzer. Using this al- specimens are a rather frequent occurrence in laboratory ternative assay configuration, the absorbance threshold for re- cording the coagulation time is dynamically increased by the practice, and the relative prevalence is described as being evaluation software of the analyzer. The threshold level depends as high as 3.3% of all of the samples afferent to a clinical 9 on the inherent absorbance and is automatically calculated by the laboratory. Hemolysis, and blood cell lysis in general, are analytic system. Plasma D-dimer was measured with the Vidas caused by biochemical, immunologic, physical, and chem- DD, a rapid, quantitative, automated enzyme-linked immunosor- ical mechanisms. In vivo blood cell lysis, which can arise bent assay with fluorescent detection, on the Mini Vidas Immu- from hereditary, acquired, and iatrogenic conditions (such noanalyzer (bioMe´rieux, Marcy l’Etoile, France). Actual reference as autoimmune hemolytic anemia, severe infections, intra- ranges were between 10.8 and 13.1 seconds for PT, 26.2 and 36.0 vascular disseminated coagulation, or transfusion reac- seconds for aPTT, 150 and 400 mg/dL for fibrinogen, and less tions), does not depend on the technique of the health- than 500 ng/mL for D-dimer. All measurements were performed care provider; thus, it is virtually unavoidable and not in duplicate within a single analytical session and the results 10 were averaged. Analytical imprecision, expressed in terms of the solvable. Conversely, in vitro blood cell lysis might be mean interassay coefficient of variation, was between 2% and 5%, prevented, because it is usually caused by inappropriate according to the manufacturers. specimen collection, handling, and processing. In the case of specimen collection, hemolysis might result from cum- Statistical Evaluation bersome or traumatic specimen collection and processing, Differences between coagulation measurements on aliquots of such as unsatisfactory phlebotomy attempts, difficulty lo- the same sample containing serial concentrations of homologous cating venous accesses, prolonged tourniquet time, wet- lysate were evaluated by paired Student t test. Statistical signif- alcohol transfer from the skin into the blood specimen, Ͻ icance was set at P .05. Percentage variations from the baseline small or fragile veins, missing the vein, syringes or but- value were further compared with the current analytical quality terfly collection devices, small-gauge needles, partial ob- specifications for desirable bias, as derived from the intraindivid- struction of catheters, vigorous tube mixing and shaking, ual and interindividual variations (Ϯ2.0% and Ϯ0.24 seconds for PT; Ϯ2.3% and Ϯ0.69 seconds for aPTT, Ϯ4.8% and Ϯ15 mg/dL or exposure to excessively hot or cold temperatures, al- 11 for fibrinogen, respectively).7 This comparison was not feasible though slow leakage may also occur. Hemolysis and for D-dimer testing, because no definitive data on its biologic blood cell lysis may not be evident until centrifugation of variability in healthy individuals are available, to our knowledge. the whole blood specimen has been performed, exposing the serum or plasma to scrutiny. RESULTS Consistent quality specimens can only result from prop- Results of our investigation are synthesized in the Fig- er training and the knowledge of the factors that can in- ure. The addition of blood cell lysates generated a consis- fluence laboratory results. Hemolysis, reflecting a more tent and dose-dependent trend toward overestimation of generalized blood cell lysis, is the most frequent reason PT (Figure, A) and D-dimer values (Figure, D), whereas for specimen rejection, as indicated by the College of 182 Arch Pathol Lab Med—Vol 130, February 2006 Hemolysis and Coagulation Testing—Lippi et al Influence of blood cell lysis on prothrombin time (PT; A), activated partial thromboplastin time (aPTT; B), fibrinogen testing (C), and dimerized plasmin fragment D testing (D-dimer; D). Differences are given in percentage (mean Ϯ SD) from the baseline sample (no lysis). The dashed horizontal lines indicate the current limits of the analytical quality specifications for desirable bias. Statistically significant differences are evaluated by Student paired t test († P Ͻ .05; ‡ P Ͻ .01).

American Pathologists Chemistry Specimen Acceptance substances from either leukocytes or , which are Q-Probes study.12 In fact, the release of hemoglobin and thought to be responsible for shortening the aPTT, as re- additional intracellular contents from erythrocytes, white ported by Garton and Larsen16 However, there is no de- blood cells, and platelets into the surrounding fluid might finitive evidence or agreement in the current literature on falsely elevate measurable levels of the same substances in this topic, to our knowledge. Among laboratory testing, serum and plasma, or might cause dilution effects, which PT, aPTT, fibrinogen, and D-dimer measurements are may compromise the reliability of laboratory testing.13 Ad- thought to be more susceptible to variations in the prean- ditionally, plasma hemoglobin might increase the optical alytical phase.17 Results of our investigation indicate that absorbance or change the blank value, producing method- a slight lysis, as low as 0.9%, might influence the reliability dependent and analyte concentration–dependent spectro- of some coagulation testing. We are aware that our inves- photometric interference with common laboratory assays.9 tigation has 2 major limitations. First, we reproduced in It is traditionally accepted that both in vivo and, more vitro blood cell lysis by freezing blood specimens at commonly, in vitro blood cell lysis can cause preanalytical Ϫ70ЊC for up to 24 hours. In addition to lysed erythro- variability. Although the amount of interference will de- cytes, hemolyzed specimens encountered in the laboratory pend on the degree of lysis and on the specificity of the practice usually contain destroyed or fragmented platelets method being used, several laboratory results can be af- and leukocytes. Therefore, our method seems to be a suit- fected, especially potassium, sodium, calcium, magne- able surrogate, although it is not representative of all of sium, bilirubin, , total protein, aldolase, amy- the possible events that can induce whole blood lysis in lase, lactate dehydrogenase, aspartate aminotransferase, laboratory practice, especially those represented by trou- alanine aminotransferase, phosphorus, alkaline phospha- blesome blood collection. Second, we evaluated interfer- tase, acid phosphatase, ␥-glutamyl transpeptidase, folate, ence on testing of single reagents and with defined ana- and iron measurements.13–15 It is traditionally known that lyzers; therefore, our results might not be universally re- routine coagulation testing might also be influenced by producible or transferable to other testing systems. This is blood cell lysis. The interference is not necessarily only particularly true for analyzers that measure the extent of caused by hemoglobin, because many substances are re- hemolysis by comparing the absorption of samples at 2 leased from blood cells and could influence coagulation wavelengths, or allow performance of coagulation assays assays. In fact, it is more likely that the effects we observed at alternative wavelengths, such as 570 nm, for turbid, ic- are caused by release of intracellular and thromboplastic teric, and hemolytic samples. The light source of the Beh- Arch Pathol Lab Med—Vol 130, February 2006 Hemolysis and Coagulation Testing—Lippi et al 183 ring Coagulation System photometer is a xenon flasher of the influence of interpretations on laboratory test ordering. Arch Pathol Lab Med. 2004;128:1424–1427. lamp with broadband emission; an interference filter with 2. Tripodi A, Breukink-Engbers WG, van den Besselaar AM. Oral anticoagulant an appropriate main wavelength is swung into the beam monitoring by laboratory or near-patient testing: what a clinician should be aware of the light source to obtain light with the desired wave- of. Semin Vasc Med. 2003;3:243–254. 3. Lippi G, Guidi GC. Effect of specimen collection on routine coagulation length. The instrument does not identify hemolytic spec- assays and D-dimer measurement. Clin Chem. 2004;50:2150–2152. imens with a specific flag. However, in our experience, 4. Horsti J. Preanalytical aspects of routine coagulation measurements. Scand results obtained by conventional assays were systemati- J Clin Lab Invest. 2001;61:167–168. 5. Bartels PC, Schoorl M, van Bodegraven AA. Reduction of preanalytical er- cally flagged as questionable by the addition of a question rors due to in vitro activation of coagulation. Clin Lab. 2001;47:449–452. mark, starting from a final percentage of lysis ranging 6. NCCLS. Reference and Selected Procedures for the Quantitative Determi- from 0.5% to 0.9%. nation of Hemoglobin in Blood: Approved Standard. 3rd ed. Wayne, Pa: NCCLS; 2000. NCCLS document H15-A3. The issue of hemolysis and blood cell lysis has plagued 7. Ricos C, Alvarez V, Cava F, et al. Current databases on biologic variation: clinical laboratories and continues to be a growing con- pros, cons and progress. Scand J Clin Lab Invest. 1999;59:491–500. 8. Burns ER, Yoshikawa N. Hemolysis in serum samples drawn by emergency cern. The most frequent causes, such as sampling errors, department personnel versus laboratory phlebotomists. Lab Med. 2002;33:378– are avoided by using standardized materials and methods 380. for the preanalytical processes and by training and indi- 9. Carraro P, Servidio G, Plebani M. Hemolyzed specimens: a reason for re- jection or a clinical challenge? Clin Chem. 2000;46:306–307. vidual counseling. Although hemolyzed and unsuitable 10. Rother RP, Bell L, Hillmen P, Gladwin MT. The clinical sequelae of intra- samples are unlikely to be observed in a problem-free vascular hemolysis and extracellular plasma hemoglobin: a novel mechanism of phlebotomy activity, each laboratory should document the human disease. JAMA. 2005;293:1653–1662. 11. Kroll MH, Elin RJ. Interference with clinical laboratory analyses. Clin procedures that are influenced by blood cell lysis and to Chem. 1994;40:1996–2005. what extent they are affected. Some instruments report an 12. Jones BA, Calam RR, Howanitz PJ. Chemistry specimen acceptability: a ability to correct the results for hemolysis, but after our College of American Pathologists Q-Probe study of 453 labs. Arch Pathol Lab Med. 1997;121:19–26. 18 experience, and, at variance with other observations, we 13. Laessig RH, Hassemer DJ, Paskey TA, Schwartz TH. The effects of 0.1% do not recommend hemolysis correction in clinical prac- and 1.0% erythrocytes and hemolysis on serum chemistry values. Am J Clin Pathol. 1976;66:639–644. tice, given the wide interindividual and analytical varia- 14. Sonntag O. Haemolysis as an interference factor in clinical chemistry. J tions arising from the interference. Therefore, if hemolysis Clin Chem Clin Biochem. 1986;24:127–139. and blood cell lysis results from an in vitro cause, we 15. Yucel D, Dalva K. Effect of in vitro hemolysis on 25 common biochemical tests. Clin Chem. 1992;38:575–577. suggest that the most convenient corrective measure 16. Garton S, Larsen AE. Effect of hemolysis on the partial thromboplastin time. might be free hemoglobin quantification and sample rec- Am J Med Technol. 1972;38:408–410. ollection. 17. Wiwanitkit V. Rejection of specimens for prothrombin time and relating pre-analytical factors in blood collection. Blood Coagul . 2002;13: 371–372. References 18. Jay D, Provasek D. Characterization and mathematical correction of he- 1. Laposata ME, Laposata M, Van Cott EM, Buchner DS, Kashalo MS, Dighe molysis interference in selected Hitachi 717 assays. Clin Chem. 1993;39:1804– AS. Physician survey of a laboratory medicine interpretive service and evaluation 1810.

184 Arch Pathol Lab Med—Vol 130, February 2006 Hemolysis and Coagulation Testing—Lippi et al